Lipopolysaccharide Primes Human Macrophages for Noncanonical Inflammasome-Induced Extracellular Vesicle Secretion

Inflammasomes are a group of intracellular pattern recognition receptors enabling early detection of pathogenic infection and metabolic dysfunctions. Inflammasome-mediated responses are critical for triggering and controlling inflammation in microbial infections and other autoimmune and autoinflammatory diseases, including gouty arthritis (1, 2), atherosclerosis (3, 4), and Alzheimer’s disease (5). Canonical NLR family pyrin domain containing 3 (NLRP3) inflammasome, the best characterized inflammasome, is activated by a range of microbial compounds as well as endogenous danger signals such as extracellular ATP and crystalline structures (6–9). Upon activation, the NLRP3 receptor assembles into a catalytic protein complex consisting of NLRP3, adaptor protein ASC, and a cysteine protease caspase-1. Subsequently, caspase-1 autoactivates to process proinflammatory cytokines pro–IL-1β and pro–IL-18 into their mature forms.

LPS constitutes an important pathogen-associated molecular pattern exposed on the surface of Gram-negative bacteria. Extracellular LPS is detected by TLR4, whose engagement results in expression of inflammatory genes (most notably pro–IL-1β) via two major signaling pathways involving NF-κB and IFN regulatory factors (IRFs) (10). During the past decade a second pathway of LPS recognition, a so-called noncanonical inflammasome, has been characterized (11–17). The noncanonical inflammasome, formed by caspase-4/5 (caspase-11 in mice), recognizes the LPS intracellularly and induces proteolytic processing and activation of gasdermin D (GSDMD). Subsequently, the N-terminal fragment of GSDMD assembles in the cellular membrane-forming pores, causing cell lysis (11, 18–20). Caspase-4/5/11 and GSDMD are involved in the pathogenesis of Gram-negative bacterial infection-associated sepsis. Besides proinflammatory cell death, the caspase-4/5/11 inflammasome also activates the NLRP3 inflammasome and subsequent secretion of IL-1β and IL-18.

Recently, ninjurin-1 (NINJ1) has been identified to mediate protein secretion and plasma cell rupture in response to noncanonical inflammasome activation (21, 22). NINJ1 has been shown to be activated through a caspase-4/11– and GSDMD-dependent manner, after which it forms pores on the plasma membrane of macrophages (22).

Extracellular vesicle (EV) secretion constitutes a major form of communication between cells, in which signaling molecules such as proteins, lipids, and nucleic acids are protected from degradation (23–25). Both EV cargo and quantity are subjects of modulation upon various abnormal conditions, including cellular stress and disease, influencing intercellular communication. Alterations in EV secretion have been implicated in inflammasome-dependent inflammatory diseases such as rheumatoid arthritis (26, 27), lupus (28), and diabetes mellitus (29, 30).

We have previously shown that several canonical NLRP3 inflammasome activators induce EV-mediated protein secretion in human monocyte-derived macrophages (hMDMs) and studied the regulatory mechanisms of this phenomenon (31–35). Also, a bacterial toxin, nigericin, has been shown to induce NLRP3- and caspase-1–dependent EV release (36, 37). ATP, which is a canonical inflammasome activator, also induces robust calpain-dependent EV secretion (33), but it is independent of the NLRP3/caspase-1 pathway (33). ATP-induced EVs carry significant numbers of signaling proteins, and they have been linked to propagation of inflammation (38).

Previously we showed increased secretion of proteins in human macrophages upon LPS transfection (39). In the current study we performed a detailed proteomic characterization of noncanonical caspase-4/5 inflammasome-induced EVs and compared it with ATP-induced EV secretion from hMDMs, utilizing high-resolution quantitative liquid chromatography–tandem mass spectrometry (LC-MS/MS) combined with bioinformatics and biochemical analysis. We show that noncanonical inflammasome activation exerted by the intracellular delivery of LPS induces EV secretion in human macrophages, and that priming of the cells with LPS extracellularly robustly augments caspase-4/5–induced, but not ATP-induced, EV-mediated protein secretion. This EV secretion is entirely dependent on caspase-4, but only partially on caspase-5 and NLRP3. Moreover, the overall pattern of EV-mediated protein secretion upon noncanonical inflammasome activation significantly differs from ATP-induced vesicles, suggesting different mechanisms of protein sorting and recruitment to the secreted EVs. Several secreted proteins were associated with the noncanonical inflammasome function, including caspase-1, caspase-4, GSDMD, and NINJ1. Bioinformatic analysis of the proteomic data indicates differential involvement of numerous cell signaling pathways in modulation of EV secretion. Importantly, we identify autocrine IFN signaling induced by extracellular LPS as the key contributor in the regulation of the EV-mediated protein secretion by the noncanonical inflammasome. Applying connectivity map analysis to proteomic datasets we identify 15-deoxy-Δ^12,14-prostaglandin J₂ (PGJ₂) and parthenolide as inhibitors of noncanonical inflammasome-induced inflammation and EV secretion. Moreover, we show that Escherichia coli–derived outer membrane vesicles (OMVs), previously shown to deliver LPS to cytosol for caspase-4/5 inflammasome activation, also induce EV-mediated protein secretion from human macrophages.

ATP (3 mM in cell culture), caspase-4 inhibitor Ac-LEVD-CHO (ICH-2, 25 μM), PGJ₂ (2–5 μM), and parthenolide (3–10 μM) were obtained from Sigma-Aldrich (Merck). LPS (1 μg/ml, E. coli K12) and ultrapure LPS (2 μg/ml, E. coli K12) were purchased from InvivoGen. Lipofectamine 2000 (5 μl/ml, Thermo Fisher Scientific) was used for the transfection of LPS. E. coli–derived OMVs (10 μg/ml) were purchased from InvivoGen. NLRP3 inhibitor Cy-09 (10-25 μM) was obtained from Tocris Bioscience. Deucravacitinib (BMS-986165, 10 μM) was obtained from Selleck Chemicals. Primary Abs detecting annexin A1 (sc-12740), galectin-3 (sc-32790), alix (sc-53540), IFIT3 (sc-393512), and guanylate binding protein (GBP)1 (sc-53857) were obtained from Santa Cruz Biotechnology. Anti–caspase-4 Ab (no. 4450) was purchased from Cell Signaling Technology. Anti-CD11c (ITGAX, ab52632) was purchased from Abcam. HRP-conjugated polyclonal goat anti-rat Ab was purchased from Invitrogen (Thermo Fisher Scientific). HRP-conjugated polyclonal goat anti-rabbit and rabbit anti-mouse were purchased from Dako/Agilent Technologies.

Human macrophages were derived from leukocyte-rich buffy coats from healthy blood donors (Regional Blood Donation Service, Lodz, Poland and Finnish Red Cross Blood Service, Helsinki, Finland). Monocytes (two to four donors per experiment) were isolated and differentiated into macrophages as previously described (32, 40). Briefly, 6–7 million PBMCs/ml were seeded on cell culture plates, and the monocytes selected by adhesion were cultured in serum-free macrophage media (Life Technologies, Thermo Fisher Scientific) supplemented with 10 ng/ml GM-CSF (ImmunoTools, Friesoythe, Germany) and 50 U/ml penicillin-streptomycin at 37°C and 5% CO₂ for 6 d, polarizing the monocytes into macrophages of the acute proinflammatory M1 phenotype.

For proteomic analyses macrophages were primed with 1 μg/ml LPS for 16 h or left untreated. Subsequently, cells were washed and supplied with fresh serum-free RPMI 1640 medium (Biowest, Nuaillé, France) with antibiotics. For noncanonical inflammasome activation, macrophages were transfected with 2 μg/ml ultrapure LPS with Lipofectamine 2000 (5 μl/ml) for 6 h. This time point was chosen due to significant protein secretion with limited cell death (W. Cypryk, unpublished observations). Alternatively, the cells were stimulated with 3 mM ATP for 15 min, which has been previously shown to trigger secretion of large numbers of EVs and activate the canonical NLRP3 inflammasome (33, 41–43). In indicated functional experiments, 4-h priming was performed. E. coli–derived OMVs were incubated with the cells for 18 h. All inhibitors were added to the cells 30 min prior to stimulation.

THP-1 cells with CRISPR-Cas9 knockouts (KOs) of caspase-4 and caspase-5 were provided by Prof. S. Masters (Walter and Eliza Hall Institute of Medical Research, Melbourne, VIC, Australia). Cells were treated with doxycycline (1 μg/ml), propagated in RPMI 1640 medium supplemented with 10% FBS, antibiotics, l-glutamine, sodium pyruvate (1 mM), and 2-ME (1% v/v), and seeded in six-well culture plates at 500,000 cells/ml. Cell lysates were prepared and CRISPR-Cas9 KOs were verified by a T7 endonuclease I assay, Sanger sequencing, and immunoblotting of caspase-4 (Supplemental Fig. 1). For PCR, in-house–made primers were used (courtesy of Dr. A. Maciaszek and B. Mikołajczyk). Primer sequences were designed using National Center for Biotechnology Information Primer-BLAST (Supplemental Fig. 1A). For activation of the noncanonical inflammasome, cells were primed with LPS (1 μg/ml) for 3 h, washed, and resuspended in the RPMI 1640 medium without serum and subsequently transfected with 2 μg/ml ultrapure LPS using Lipofectamine 2000 (5 μl/ml) for 6 h. Culture medium from an equal number of cells across samples was analyzed directly by ELISA or concentrated with 10-kDa molecular mass cutoff centrifugal filter units (Millipore, Merck) for immunoblotting.

Conditioned cell culture media were cleared from cells and large cell debris by serial centrifugation at 500 × g (10 min) and 3000 × g (30 min). For gel-based analyses, medium from an equal number of cells per condition was concentrated to equal a final volume using Amicon 100-kDa molecular mass cutoff centrifugal filter units (Millipore). Subsequently, the concentrated medium was mixed with Laemmli loading buffer and upon protein denaturation analyzed on SDS-PAGE (EV-enriched fraction). For nanoparticle tracking analysis (NTA) and proteomic analyses, cell-free culture supernatants were concentrated with Amicon 100-kDa molecular mass cutoff centrifugal filter units (Millipore), and EVs were isolated by ultracentrifugation using a SW41 rotor (Beckman Coulter). Ultracentrifugation was performed at 34,000 rpm (k-factor 180, average relative centrifugal force 142,000 × g) for 2 h at 4°C. After two ultracentrifugations with an intermediate PBS wash, the EV pellet was resuspended in PBS for proteomic analysis and NTA. For proteomics, purified EVs from seven independent biological replicates were stored at −80°C before protein digestion and LC-MS/MS analysis.

For MS analyses the EVs resuspended in 100 μl of sterile PBS were mixed with 10 µl of 1% ProteaseMAX surfactant (Promega, Madison, WI) in 50 mM ammonium bicarbonate (Sigma-Aldrich). Samples were vortexed for 1 min, then heated at 95°C for 5 min before sonication for 20 min in a water bath. Cysteines were reduced with 10 mM DTT (Sigma-Aldrich, Winston Park, ON, Canada) at 56°C for 30 min and alkylated with 14 mM iodoacetamide (Sigma-Aldrich, St. Louis, MO) for 20 min at room temperature, and proteins were digested by trypsin (Promega, Madison, WI) at 37°C overnight. The resulting peptides were purified and concentrated using C18 microcolumns (C18 Empore extraction disk, Varian, St. Paul, MN).

The peptide samples were analyzed by LC-MS/MS using an Easy nLC1000 connected to a Q Exactive HF (Thermo Fisher Scientific) with a 60-min separation gradient. MS raw files were submitted to MaxQuant software version 1.6.1.0 (44) for protein identification and label-free quantification. Parameters were set as follows: carbamidomethylation as fixed modification and protein N-acetylation and methionine oxidation as variable modifications. Trypsin without a proline restriction enzyme option was used, with two miscleavages allowed. Minimal unique peptides were set to 1, and the false discovery rate allowed was 0.01 (1%) for peptide and protein identification. A database search was done against UniProt human (version September 2018). Perseus (version 1.6.1.3) (45, 46) was used for additional data processing and statistical analysis. In Perseus, the intensity values were log₁₀ transformed, quantification data were filtered requiring a minimum of three valid values in at least one group, and missing values were imputed from normal distribution. For statistical comparison, a paired Student t test with p < 0.05 was used.

A Gene Ontology (GO) search was performed using g:Profiler (47). IFN-regulated proteins were identified using the Interferome database (version 2.01) (48) with filtering for IFN type I gene regulation threshold set to 4 in hMDMs. Gene Set Enrichment Analysis (GSEA) was performed with GSVA R package (49) and the GSEA 4.1.0 desktop application (50) using curated gene sets of canonical pathways derived from the Reactome pathway database (51–53). For comparisons between experimental conditions, a preranked GSEA with a t score for dependent observations was used. A leading edge analysis (LEA) was performed to identify the core enrichment proteins in EVs after macrophage conditioning. Connectivity map analysis (54, 55) was performed based on LEA results to determine drugs capable of inhibiting LPS transfection-dependent EV secretion.

NTA was performed using NanoSight NTA (Malvern Panalytical) with software version 3.4. Vesicles were analyzed at a detection threshold of 5 and a camera level of 14 at 25°C in triplicates (60 s per measurement). Samples containing EVs were diluted in filtered PBS for analysis.

Cellular RNA was isolated using an RNeasy plus kit (Qiagen, Düsseldorf, Germany) according to the manufacturer’s instructions. cDNA was synthesized from 1 μg of RNA using an iScript cDNA synthesis kit (Bio-Rad, Hercules, CA). Quantitative RT-PCR was performed with 5 ng of cDNA using 5× HOT FIREPol EvaGreen (Medibena, Vienna, Austria) in an RT-PCR LightCycler 96 (Roche). The reference gene was RPLP0.

SDS-PAGE was carried out using TGX gels (Bio-Rad). Silver staining (56) was used for visualization of proteins in gel. For immunoblotting, proteins were transferred to Immobilon-P transfer membranes (Merck Millipore) and probed with primary Abs overnight. Upon incubation with secondary Abs the membranes were visualized using the Uvitec Alliance Q9-365 advanced imaging system (BioSPX) or ChemiDoc Imaging System (Bio-Rad). Sample loading for silver stain and immunoblot analyses of enriched EV fraction and purified EV was normalized against the number of cells stimulated (i.e., each sample contained proteins secreted from equal number of cells per condition). Therefore, higher EV-mediated protein secretion is reflected by a stronger signal shown in silver stain and immunoblots throughout this article.

Human DuoSet ELISA kits for detection of human IL-1β and human IL-18 were purchased from R&D Systems. All analyses were performed on cell-free conditioned culture medium according to the manufacturer’s instructions.

Conditioned cell culture media were centrifuged briefly to remove cells and cell debris and analyzed with a Pierce lactate dehydrogenase (LDH) cytotoxicity assay kit (Thermo Scientific) or a cytotoxicity detection kit (Roche) according to the manufacturers’ instructions.

Comparisons between proteomic quantitative data were tested using either a t test for dependent variables or repeated measures ANOVA. For sample clustering, unsupervised (principal component analysis [PCA]) and supervised (partial least squares discriminant analysis [PLS-DA]) methods were applied. GSEA and preranked GSEA were performed using gene set permutation and appropriate statistics as described above. Gene set clustering and selection were based on t-distributed stochastic neighbor embedding and repeated measures ANOVA. The similarity between selected gene sets was based on enrichment in tested comparisons (hierarchical clustering with average clustering method and Ward’s distance measure) and genes (hierarchical clustering with average clustering method and occurrence frequency measure). Comparisons between other variables were tested using appropriate statistics depending on normality as tested with the Shapiro–Wilk test. For continuous variables, either ANOVA and a Student t test or Kruskal–Wallis ANOVA and a Mann–Whitney U test were applied. For nominal variables, a χ² test, a χ² with a Yates correction for continuity, or a two-tailed Fisher’s exact test was applied, depending on the number of observations. A comparison was determined significant with a p value of <0.05 and false discovery rate of <0.15. Multiple testing was controlled using the Benjamini–Hochberg correction. Quantitative RT-PCR, ELISA, and LDH data of the fold changes in gene expression were tested using a paired Student t test for matching samples (n ≥ 3 in each experiment, independent experiments).

The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (57, 58) partner repository with the dataset identifier PXD033002.

Supplemental Fig. 1 shows validation of investigated THP-1 KO cells. Supplemental Fig. 2 presents SDS-PAGE–based analyses and cytotoxicity data showing the effect of Lipofectamine 2000–only and extracellular LPS-only treatments on EV secretion (A–C) and the effect of LPS priming on ATP-induced EV secretion (D). Supplemental Table I contains the list of all proteins identified in seven biological replicates together with their total intensities in different samples (A) and the quantitative comparison of the abundance of proteins identified in seven biological replicates of the experiment (B). Supplemental Table II contains T-scores from donor-paired comparisons of proteins in EVs from macrophages activated in all combinations of four experimental conditions: (A) t-LPS versus ctrl, prim+t-LPS versus ctrl, ATP versus ctrl, t-LPS versus prim+t-LPS, prim+t-LPS versus ATP, and t-LPS versus ATP. (B) GSEA results for Reactome C2 canonical pathways in comparisons: global (repeated measures ANOVA), t-LPS versus ctrl, prim+t-LPS versus ctrl, ATP versus ctrl, t-LPS versus prim+t-LPS, prim+t-LPS versus ATP, and t-LPS versus ATP. Supplemental Table III includes data of LEA analysis (A) and CMap analysis of prim+t-LPS versus ATP (B).

All blood donors gave written permission for using their donated blood for research purposes during donation in the Regional Blood Donation Service, Lodz, Poland or the Finnish Red Cross Blood Service, Helsinki, Finland, maintaining their anonymity. Neither the identities of the donors nor any information concerning their sex, age, or health was disclosed to the authors at any stage. The buffy coats were HIV-, hepatitis B virus–, and hepatitis C virus–negative.

We have previously shown an increase in protein secretion upon noncanonical inflammasome activation in human macrophages (39). In this study, we performed a detailed quantitative proteomic analysis of purified EVs secreted by untreated and LPS-primed hMDMs upon ultrapure LPS transfection, which activates caspase-4 (14). To obtain a broader picture of EV secretion, we have also analyzed EV secretion upon treatment with ATP, which also activates the canonical NLRP3 inflammasome (7). Millimolar ATP induces robust and fast secretion of EVs in human macrophages, which is independent of the NLRP3/caspase-1 pathway (33, 41–43).

To assess the induction of EV secretion, we isolated total EVs by ultracentrifugation from equal numbers of control macrophages (ctrl), untreated macrophages and LPS-primed macrophages transfected with LPS for noncanonical caspase-4/5 inflammasome activation (t-LPS and prim+t-LPS, respectively), and from ATP-pulsed macrophages. Silver stain and immunoblot of vesicular membrane marker integrin α_X revealed that LPS transfection markedly increases the EV secretion. Neither mock transfection with Lipofectamine alone nor priming via extracellular LPS alone had any impact on EV secretion (Supplemental Fig. 2A–C). Interestingly, the induction of EV secretion exerted by LPS transfection was substantially enhanced when macrophages were activated with extracellular LPS to activate TLR4 signaling before LPS transfection (Fig. 1A). As observed before (33), ATP treatment activated robust EV secretion (Fig. 1A); however, ATP-induced EV secretion was not enhanced by LPS priming (Supplemental Fig. 2D).

We used NTA to quantify the secreted EVs and assess their size distribution. The analysis confirmed that upon caspase-4/5 inflammasome activation the EV secretion was increased ∼3-fold compared with control macrophages. Priming preceding the LPS transfection further increased this ratio to 18 (Fig. 1B). As expected, ATP treatment resulted in the largest, ∼80-fold increase in numbers of secreted EVs compared with unstimulated cells (Fig. 1B). Size distribution analysis showed that most secreted vesicles in all samples were <200 nm, regardless of the stimulus, which is a size corresponding to small EVs (Fig. 1C). However, it is evident that LPS transfection also induces release of vesicles >200 nm (Fig. 1C).

Taken together, our results indicate that noncanonical inflammasome activation triggers EV-mediated protein secretion from human macrophages, which is dramatically enhanced by LPS priming. To study the mechanism of this phenomenon, we analyzed the extracellular LPS-induced expression of noncanonical inflammasome components as well as proteins that potentially regulate caspase-4/5–mediated protein secretion, including GSDMD and NINJ1, using quantitative RT-PCR. As expected, a strong upregulation of IL1B gene expression was seen already at 2 h after extracellular LPS priming (Fig. 1D). LPS priming had little effect on CASP1, CASP4, and GSDMD gene expression, which were induced <2-fold throughout LPS priming (Fig. 1D). However, priming with extracellular LPS significantly enhanced CASP5 and NINJ1 gene expression with ∼5- to 10-fold increased mRNA levels seen between 2 and 6 h postactivation (Fig. 1D). These results suggest that extracellular detection of LPS may enhance caspase-4/5–induced EV-mediated protein secretion by upregulation of caspase-5 and NINJ1 expression.

Both caspase-4 and caspase-5 directly bind cytosolic LPS (14). Caspase-4 autoactivates upon oligomerization, which constitutes the first step of noncanonical inflammasome activation, whereas the role of caspase-5 has remained ambiguous. To investigate whether caspase-4 is required for activation of EV secretion in human macrophages, we pretreated the cells with caspase-4 inhibitor ICH-2 (Ac-LEVD) before LPS transfection. Inhibition of caspase-4 markedly reduced IL-18 secretion and cell death manifested by the release of LDH both in untreated and LPS-primed macrophages (Fig. 2A). Likewise, caspase-4 inhibitor reduced IL-1β secretion in primed and LPS-transfected cells (Fig. 2A). Importantly, caspase-4 inhibition also led to significant downregulation of EV-associated protein secretion, confirming that LPS transfection–induced EV secretion requires functional caspase-4 (Fig. 2A).

Baker et al. (59, 60) previously identified differential roles of caspase-4 and caspase-5 in sensing intracellular LPS in THP-1 monocytes. Supporting this notion, a recent work by Wang et al. (61) clarified that caspase-5 is required for liberation of LPS from endosomes and exposing it for the interaction with caspase-4. We hypothesized that a similar discrepancy between caspase-4 and caspase-5 may influence EV secretion. To investigate this hypothesis, caspase-4 and caspase-5 KO THP-1 cells (59, 60) were left untreated or subjected to extracellular LPS priming and subsequent LPS transfection. Disruption of caspase-4 and caspase-5 genes in respective clones was confirmed by a T7 endonuclease 1 assay (Supplemental Fig. 1A) and Sanger sequencing (Supplemental Fig. 1B, 1C). Caspase-4 immunoblot confirmed the lack of caspase-4 expression in caspase-4 KO cells, but not in caspase-5 KO cells (Supplemental Fig. 1D). Similar to Baker et al. (59), we observed that caspase-4 KO cells secreted almost 80% less IL-1β than did their wild-type (WT) counterparts, whereas caspase-5 KO cells secreted only ∼40% less IL-1β than did WT THP-1 cells (Fig. 2B, left panel). To investigate the EV-mediated protein secretion, we analyzed the levels of vesicular marker alix from EV-enriched cell culture media from control and LPS-primed and LPS-transfected cells. EV secretion was markedly reduced in caspase-4–deficient THP-1 cells as compared with WT, whereas less pronounced reduction was evident in caspase-5 KO cells (Fig. 2B, right panel). Taken together, these results demonstrate that caspase-4 and caspase-5 may differentially regulate EV secretion induced by noncanonical inflammasome activation.

Our previous studies indicated that several canonical NLRP3 activators including β-glucans, ATP, influenza A virus, and crystalline substances induce EV secretion in human macrophages (31–34). NLRP3 inflammasome is also activated downstream of noncanonical caspase-4/5 activation (62, 63) and is required for processing of proinflammatory cytokines IL-1β and IL-18 upon cytosolic LPS sensing (11, 64). To study whether NLRP3 activity contributes to induction of EV-mediated protein secretion, we pretreated untreated and LPS-primed macrophages with the selective inhibitor of NLRP3 (Cy-09) (65) before transfecting the cells with LPS. Inhibition of NLRP3 resulted in 35–50% reduction in IL-18 secretion from both untreated and LPS-primed cells (Fig. 2C). Similarly, IL-1β secretion in primed and transfected cells was inhibited by Cy-09 (Fig. 2C). To analyze EV secretion, we performed silver staining and immunoblot analyses of the EV-enriched fraction of the culture media, revealing that inhibition of NLRP3 results in partial, dose-dependent reduction of the EV-mediated protein secretion. These data demonstrated that NLRP3 inflammasome activity downstream of caspase-4 contributes to the secretion of EVs in human macrophages. Importantly, this effect was observed both in untreated and LPS-primed cells (Fig. 2C).

Taken together, our data show that LPS transfection-induced EV secretion is a completely caspase-4– and partially NLRP3-dependent phenomenon, and it is significantly enhanced by LPS priming in human macrophages.

To characterize EV-mediated protein secretion during noncanonical inflammasome activation in detail and to compare their proteome with ATP-induced vesicles, we next performed label-free quantitative proteomic analyses of the EVs isolated from control, ATP-stimulated, and LPS-transfected (both untreated and LPS-primed) macrophages (Fig. 3A). We identified and quantified a total of 2890 proteins (Supplemental Table I). The quantitative proteomic data showed that EV-mediated secretion of >1100 proteins was significantly increased by LPS transfection in macrophages as compared with unstimulated cells. In total, 2384 proteins were more abundant in EVs after LPS priming and LPS transfection compared with EVs from control cells, and EV-mediated secretion of >2600 proteins increased after ATP stimulation (paired Student t test, p < 0.05). To verify the nature of the vesicles we isolated, we looked for the presence of canonical small EV- and microvesicle-associated marker proteins (66). As expected, both large and small EVs contributed to the observed proteome of analyzed vesicles (Table I). In addition to inflammatory cytokines IL-1β and IL-18, we also identified numerous proteins directly or indirectly associated with canonical and noncanonical inflammasome function, including IL-1β, IL-18, caspase-1, caspase-4, GSDMD, and apoptosis-associated speck-like containing a CARD (ASC), NLRP3 activity regulating NEK7 (67), and recently described membrane rupture-associated protein NINJ1 (22) (Table II). Interestingly, priming of the cells with LPS extracellularly prior to LPS transfection strongly enhanced caspase-4/5–induced secretion of GSDMD and NINJ1.

Next, we analyzed the enrichment of GO terms for all identified proteins. Among GO molecular function terms, adhesion and protein binding were enriched (Fig. 3B). This is in agreement with our and others’ earlier observations and is associated with high abundance of membrane-associated proteins in the isolated EVs, especially in the small EVs (32, 68). We also observed enrichment of terms associated with exocytosis and vesicle-associated terms (GO:0031982, vesicle; GO:0043230, extracellular organelle; GO:0070062, extracellular exosome; and GO:1903561, extracellular vesicle). Importantly, we also observed enrichment of immune response–related terms (GO:0002275, myeloid cell activation involved in immune response; GO:0002376, immune system process; and GO:0002366, leukocyte activation involved in immune response), as expected in inflammatory response activation (Fig. 3B).

We then proceeded with an in-depth analysis of quantitative proteomic data to better understand how different stimuli (LPS transfection with and without priming, and ATP) activate EV secretion from human macrophages. PCA of protein identification and quantification data indicated good clustering of the same stimulus type samples across biological replicates. To focus on intersample rather than intrasample variation in EV protein content, we used sample-normalized expression of the 2500 most variable identified proteins. The ATP cluster was clearly separate from LPS transfection only and LPS priming followed by LPS transfection samples, underlining the differences in the regulation of the proteomic landscape of EVs between studied stimuli (Fig. 3C). Moreover, considering protein expression patterns, LPS priming plus LPS transfection samples were focused as a subset of overall transfected samples, indicating similarities between these experimental conditions. Importantly, however, note that samples after priming and transfection demonstrated much higher overall secretion of EV-associated proteins compared with LPS-transfected cells only.

Next, we performed PLS-DA to identify proteins whose EV-mediated secretion most significantly drives the overall difference between the stimuli. Altogether, 25 key cluster-separating proteins were determined (Fig. 3D). Importantly, secretion of three of these proteins, that is, galectin-3–binding protein (LGALS3BP), mitochondrial superoxide dismutase 2 (SOD2), and TNFAIP3-interacting protein 1 (TNIP1), was induced significantly more under noncanonical inflammasome activation in primed cells (Fig. 3E) than in ATP-pulsed cells, which is likely associated with TLR4-induced overexpression of these proteins (69, 70). In addition to these, 15 other proteins were more abundantly secreted in these conditions than in ATP (Fig. 3E).

Taken together, our analysis reveals significant changes of the proteome of secreted EVs and underlines the differential protein recruitment to EVs as a result of noncanonical inflammasome activation and ATP treatment, probably due to the existence of different, stimulus-specific regulatory mechanisms.

To further investigate the impact of macrophage activation on EV proteomic signatures observed, we performed GSEA of the proteomic data (Supplemental Table II). To minimize interspecimen variation, we used a t test for dependent variables and preranked GSEA. This provided us with sufficient information to determine pathways whose activation patterns are dependent on noncanonical inflammasome activation in untreated and LPS-primed cells as well as in ATP-pulsed cells (Fig. 4A). We identified several pathways involved in EV biogenesis (e.g., Rho GTPases, actin cytoskeleton remodeling [71] and Rab proteins known to regulate endosomal traffic [72]), as well as pathways linked to inflammatory process (IL-1 family signaling or MAP2K and MAPK activation, (Fig. 4A). Moreover, contribution of these pathways to the observed EV proteome was predicted to be dramatically different in noncanonical inflammasome activation and in ATP-induced EVs (Fig. 4A). We also observed that extracellular LPS priming preceding LPS transfection significantly modulates the pathway activation patterns as compared with LPS-transfection only (Fig. 4A). Interestingly, we observed a striking role of IFN signaling in regulation of vesiculation in primed and transfected cells (Fig. 4A). Extracellular binding of LPS to TLR4 activates two major signaling pathways: the inflammatory gene program activated by NF-κB and AP-1, and the IFN gene program activated by IRFs (73), resulting in secretion of type I IFN. The IFN-sensing through Toll/IL-1R domain–containing adapter inducing IFN-β (TRIF) was shown to license the caspase-4 for noncanonical inflammasome activation (62). We were therefore interested whether the IFN pathway influences the noncanonical inflammasome-mediated EV secretion. To analyze the potential effect of the TLR4-induced IFN signaling impact on EV secretion, we analyzed representation of IFN-stimulated genes (ISGs) in the proteomic datasets. All proteins with statistically significant differences in LPS priming plus LPS transfection and ATP samples as compared with control samples were analyzed by the Interferome database (48). This analysis revealed 44 proteins known to be induced at least 4-fold by type I IFN in hMDMs (Fig. 4B). To account for overall bias in total EV-mediated protein secretion between samples, we ranked the proteins by fold secretion compared with control samples and analyzed the top 10% of them. Interestingly, in EVs isolated from LPS-primed and transfected cells, ISGs constituted 13 of the 10% top upregulated proteins, including the most induced IFIT3 and Mx1, whereas only 1 of the 10% top upregulated proteins in ATP-induced EVs were dependent on IFN signal (Fig. 4B). None of the ISGs was represented in the top 10% of upregulated proteins in LPS-transfected only macrophage-derived EVs. This analysis suggested that autocrine IFN signal induced during priming with LPS significantly influences the EV proteome by recruiting IFN-induced proteins for secretion.

To directly investigate the role of IFN signaling in EV secretion we inhibited type I IFN signaling with selective tyrosine kinase 2 (TYK2) inhibitor BMS-986165 (74, 75) before LPS priming and subsequent LPS transfection. As demonstrated in (Fig. 4C, BMS-986165 did not reduce the EV protein secretion triggered by LPS transfection in untreated cells, but it significantly reduced EV secretion following priming and transfection. It also markedly reduced the secretion of IFN-induced proteins IFIT3 and GBP1 from primed and transfected cells (Fig. 4C). Taken together, these data demonstrate that TLR4-induced IFN signaling enhances EV secretion and recruitment of ISGs to EV upon noncanonical inflammasome activation.

LEA was performed to recognize major contributors to differences in EV secretion between LPS priming followed by transfection versus ATP-treated macrophages. Proteins consistently present in significantly regulated pathways (GSEA, Supplemental Table II) and differentially expressed between experimental conditions were defined as key proteins driving the observed proteomic changes. Within LEA-defined genes some differentially secreted proteins constituted protein complexes, such as 14-3-3 proteins (YWHAZ, YWHAE, YWHAQ, YWHAH, YWHAG and YWHAB), proteasomes (PSMA1, PSMA2, PSMA3, PSMA5, PSMA6, PSMB7), and other complexes (PRKCD, PARK7, PPIA, ITCH, TSG101, S100A9, details in Supplemental Table III). Thus, leading-edge proteins were not only coexpressed, but they often colocalized as members of specific molecular machinery. To investigate differential signaling governing EV-mediated protein secretion induced by LPS priming followed by transfection and ATP treatment, we applied connectivity map analysis (Broad Institute, Connectivity Map build 02) (54, 55, 76) on the list of genes recognized as major contributors (LEA). Connectivity map analysis was applied only to the major contributors recognized by LEA to eliminate possible variance not associated with the investigated comparison of macrophages activation. Through enrichment similarity and specificity scores, we determined the top 2 from a list of 14 substances predicted to result in similar or opposite effects to 1 observed in comparing EV proteome differences between LPS-primed and transfected versus ATP-pulsed macrophages (Fig. 5A). Parthenolide and PGJ₂ were selected for experimental validation in the context of their influence on EV-mediated secretion upon noncanonical inflammasome activation (Fig. 5A). Of note, both parthenolide and PGJ₂ have previously been found to interfere with NF-κB and NLRP3 inflammasome signaling (77–79). Pretreatment of cells with PGJ₂ significantly reduced EV secretion induced by LPS transfection from both untreated and primed macrophages and decreased IL-18 secretion in a dose-dependent manner (Fig. 5B). Similar results were obtained for parthenolide (Fig. 5C). Thus, we identified and experimentally verified PGJ₂ and parthenolide as inhibitors of inflammatory activity and EV secretion induced by the noncanonical inflammasome.

Gram-negative bacterial OMVs have been shown to transport LPS for intracellular recognition by caspase-4/11, resulting in the induction of proinflammatory cytokine secretion and pyroptosis (16). To verify whether bacterial OMVs, similarly to LPS transfection, induce EV secretion, we stimulated untreated and LPS-primed macrophages with OMVs derived from E.coli. As expected, OMVs induced the noncanonical inflammasome as determined by IL-1β secretion (Fig. 6A). OMVs also activated robust EV-associated protein secretion compared with control cells (Fig. 6B). Importantly, bacterial OMVs were sufficient to induce IL-1β secretion and EV release without extracellular LPS priming; however, ELISA and immunoblot analysis revealed that LPS priming preceding OMV stimulation modestly augments both the inflammatory response and EV-mediated protein secretion (integrin α_X and galectin-3, (Fig. 6B). This is in agreement with previous data showing the capability of OMVs to both prime and activate the inflammasome (80). Thus, similar to LPS transfection, also the biologically relevant noncanonical inflammasome activation by uptake of bacterial OMVs induces EV secretion.

Inflammasome activation is a typical response to bacterial and viral infections as well as diseases involving a sterile proinflammatory trigger (e.g., crystalline substances). In the past years it has become clear that Gram-negative pathogenic bacteria are capable of activating immune cells via a novel, caspase-4/5/11–dependent pathway, which has been shown to trigger pyroptosis via GSDMD and activate the NLRP3 inflammasome, resulting in processing and secretion of IL-1β and IL-18 (18–20). Functionally, this pathway has been shown to play essential role in sepsis, where large amounts of LPS accumulate in the bloodstream due to the presence of live bacteria and the OMVs they secrete (16, 81).

We have previously shown that the noncanonical caspase-4/5 inflammasome activates protein secretion in human macrophages (39). In the current study we focused on EVs purified by ultracentrifugation and performed in-depth proteomics combined with bioinformatics analysis of purified EVs secreted during this process. This approach allowed us to obtain a much more accurate proteome of EVs. EV secretion induced by LPS transfection was entirely dependent on caspase-4 both in human primary macrophages as well as in monocytic THP-1 cells (Fig. 2A, 2B). Interestingly, we also uncovered partial dependence of EV secretion on caspase-5, underlining apparent differences in the function of these related caspases (Fig. 2B). In a recent study, Wang et al. (61) showed that rather than exerting a proinflammatory effect, caspase-5, along with sorting nexin 10 (SNX10), is involved in promoting cytosolic exposure of LPS and its interaction with caspase-4. Our data suggest that the same mechanism may be involved in regulation of EV secretion by a noncanonical inflammasome, where caspase-4 is critical and caspase-5 works as an auxiliary amplifier of EV secretion.

ATP induces fast and robust EV secretion in human macrophages due to activation of P2X purinoceptor 7 (P2X7) ion channel receptor, which results in disruption of potassium/calcium equilibrium and activation of the canonical NLRP3 inflammasome (7, 82). We and others have shown that calcium mobilization triggered by P2X7 activation is a driving factor for EV release (33, 83). Moreover, millimolar ATP-induced vesiculation in human macrophages is a rapid event and does not depend on de novo protein expression (84). Our data confirm these observations, and we used the ATP-induced proteome as a reference proteomic landscape for the noncanonical inflammasome-induced EV proteome (Fig. 3C, 3D). Our results indicate that ATP and the noncanonical inflammasome induce protein trafficking to EVs through completely different mechanisms, likely leading to secretion of EVs with different biological properties and functions. Further studies are needed to identify factors that are involved in protein transport to EVs following ATP stimulation and noncanonical inflammasome activation.

We showed that extracellular LPS priming significantly sensitizes the macrophages to caspase-4/5 activation upon LPS transfection, and this is associated with a major increase of EV secretion (Fig. 2C). This phenomenon is particularly interesting, as extracellular LPS activating TLR4 does not induce robust EV secretion or cell death in hMDMs (Supplemental Fig. 2A, 2B and Ref. 31). Our data indicate that noncanonical inflammasome activation induces EV secretion that is only partially dependent on NLRP3 function both in naive as well as in primed macrophages (Fig. 2C). Therefore, we now provide evidence that extracellular and intracellular LPS detection pathways act in concert to induce secretion of EV, which is not the case with ATP-induced EV secretion (Supplemental Fig. 2D). Interestingly, we saw a robust upregulation of caspase-5 and NINJ1 mRNA expression in LPS-primed macrophages, which may explain the enhancement effect of extracellular LPS exposure on the EV secretion in response to noncanonical inflammasome activation. It is likely that LPS-induced expression of caspase-5 provides an additional caspase-5 inflammasome platform augmenting the inflammatory response and driving the secretion of EVs. Indeed, caspase-5 has been previously found to be inducible by LPS (85). Extracellular LPS-mediated overexpression of NINJ1, but not GSDMD, may indicate a possible balance shift between these two cell death effectors that has consequences in LPS transfection-induced EV secretion. A detailed investigation of the roles of terminal inflammasome pathway components may in the future provide insight into the regulatory mechanisms driving the EV secretion in inflammation.

Our proteomic analysis identified several inflammasome components and inflammasome-related proteins secreted on EVs from macrophages in response to noncanonical inflammasome activation (Table II). It has been previously shown that ASC specks are secreted in response to NLRP3 inflammasome activation. They function as danger signals, retaining capacity for caspase-1 activation (86, 87). ASC specks are readily taken up by macrophages and they have a functional role in recipient cells (86). Zhang et al. (36) identified ASC on all types of exosomes released by both untreated and inflammatory mouse macrophages. However, this ASC release was only slightly increased in response to NLRP3 activation. Our data show that ASC is secreted in EVs from human macrophages also in response to noncanonical inflammasome activation, and its secretion is strongly augmented by LPS priming (Table II). This suggests the possibility that ASC specks are secreted in association with the EVs also driven by the noncanonical caspase-4/5 activity.

Our data indicated that both canonical and noncanonical inflammasome components are secreted in the EVs (Table II). Interestingly, these included recently described effector molecules (NINJ1, GSDMD, and mixed lineage kinase domain-like protein [MLKL]) that trigger lytic cell death upon canonical and noncanonical inflammasome activation. Their EV-mediated secretion significantly rises following noncanonical inflammasome activation in LPS-primed cells. It remains to be elucidated whether EV-associated GSDMD, MLKL, and NINJ1 could be taken up and processed in the recipient cells to assemble membrane pores and exert cell lysis.

Recently a thorough analysis confirmed our previous results showing that canonical NLRP3 inflammasome activators indeed trigger a robust caspase-1 and NLRP3-dependent EV release (37). Moreover, the inflammasome activation strongly affects the RNA content of secreted EVs, equipping them with IFN-β mRNA and possibly other factors that actively stimulate type I IFN responses in recipient macrophages (37). Our GSEA analysis attributed IFN signaling to the regulation of the EV proteome of LPS-primed cells. IFN signaling was previously identified as a critical regulator of cytosolic LPS sensing through upregulation of IRF1/IRF2 (88) and the GBP family (89). Type I IFNs are expressed and secreted upon TLR4 activation by extracellular LPS. This in turn activates an autocrine IFN signaling via the IFN-α/β receptor and subsequent activation of ISG expression via TYK2 (90). TYK2 has also recently been found to be required for caspase-11 and caspase-5 expression upon TLR4 activation via the type I IFN signal (91). We now show that noncanonical inflammasome activation in LPS-primed human primary macrophages leads to dramatic induction of EV release and enrichment of these EVs with many ISGs, including GBP and IFIT family proteins (Fig. 4, Table II). Enrichment of ISGs in EVs may result in an IFN-dependent response in recipient cells (92). Noncanonical inflammasome-induced secretion of GBP1 and IFIT3 proteins was significantly reduced by the TYK2 kinase inhibitor BMS-986165 in LPS-primed macrophages (Fig. 4C). Our findings underline that an autocrine, TLR4-induced IFN signal, apart from the previously described effect on caspase-4/5 expression and function (15, 62, 91, 93), is also attributed to the noncanonical inflammasome-mediated induction of EV secretion from LPS-primed cells.

To further explore the critical role of noncanonical inflammasome activation and associated EV secretion, we employed connectivity map analysis to identify pharmacological inhibitors that may affect the EV-mediated protein secretion exerted by LPS transfection. We identified two potential inhibitors of inflammation and EV-mediated protein secretion: PGJ₂ and parthenolide. PGJ₂, which is a high-affinity ligand for the peroxisome proliferator-activated receptor γ, has previously been found to inhibit NF-κB signaling (77) and NLRP1/NLRP3 inflammasomes (79). Similarly, parthenolide has been shown to inhibit the NLRP3 inflammasome by blocking its ATPase activity and inhibiting caspase-1 (78). Although these previous studies suggested that anti-inflammatory properties of parthenolide and PGJ₂ are associated with the canonical NF-κB/NLRP3 axis, our data now show that inhibition of constitutively expressed IL-18 release and EV secretion by these two compounds occurs also in unprimed macrophages, where canonical NLRP3 activity is low. It will be important to further determine whether PGJ₂ and parthenolide can directly inhibit the noncanonical inflammasome by targeting caspase-4/5 or other effector molecules (e.g., GSDMD). If so, these therapeutics could prove useful in the treatment of acute bacterial sepsis, where large amounts of LPS associated with living bacteria and OMVs that they secrete get access to the bloodstream.

In the current study we also report that bacterial OMVs, acting as functional vehicles for intracellular LPS delivery (16), have, analogous to LPS, transfection capacity to induce EV secretion (Fig. 6). This may significantly affect the regulation of EV-mediated intercellular communication in septic shock, which may occur in case of the presence of large numbers of OMVs in the circulation (94). It is highly likely that caspase-4 activity is also driving the EV release upon OMV sensing.

In conclusion, our in-depth proteomic analysis of EV secretion following noncanonical inflammasome activation demonstrates that these extracellular and intracellular LPS recognition pathways synergistically activate the EV secretion in human macrophages. Proteins whose expression is induced during extracellular LPS priming that could contribute to enhanced EV secretion included caspase-5, NINJ1, and type I IFNs. Furthermore, proteins secreted in these inflammatory macrophage-derived EVs included several proteins that could trigger cell lysis and/or membrane rupture in recipients cells. Future in vivo studies are needed to elucidate the possible biological roles of EVs secreted upon noncanonical inflammasome activation related to bacterial infections.

We thank Prof. Seth L. Masters (Walter and Eliza Hall Institute of Medical Research, Melbourne, VIC, Australia) for providing us with THP-1 KO cell lines. We thank Dr. Rafał Madaj (Centre of Molecular and Macromolecular Studies, Polish Academy of Sciences [CMMS PAS]) for assistance in data visualization. Dr. Anna Maciaszek and Barbara Mikołajczyk (CMMS PAS) are acknowledged for synthesis and purification of PCR primers.

The authors have no financial conflicts of interest.